An improved manufacturing process for the antimalaria drug coartem. Part I

Article abstract of DOI:10.1021/op0602425

Artemisinin and its derivatives, such as artemether, are highly sensitive compounds, which require careful optimized production processes for their manufacture. Due to robustness issues, the manufacturing procedure of the reduction of artemisinin with potassium borohydride to dihydroartemisinin was re-investigated. The most important factor for obtaining optimal yields is to ensure low levels of contamination of potassium hydroxide in potassium borohydride. Application of a lower reaction temperature, fast addition rate of potassium borohydride, and careful control of the pH during the quench with acid are further important parameters in guaranteeing a robust process. In the redesign of the conversion of dihydroartemisinin to artemether, the yield was increased, and dichloromethane was replaced by the ecologically friendlier methyl acetate. A robust manufacturing process for artemether is now at hand, allowing the production of this important medicine reliably and in good quality and yield. & 2007 American Chemical Society.

Full text of DOI:10.1021/op0602425

Organic Process Research & Development 2007, 11, 336−340
An Improved Manufacturing Process for the Antimalaria Drug Coartem. Part I
Matthias Boehm, Peter C. Fuenfschilling,* Matthias Krieger, Ernst Kuesters, and Fritz Struber
NoVartis Pharma AG, Chemical and Analytical DeVelopment and Chemical Operations, CH-4002 Basel, Switzerland.
Abstract:
Artemisinin and its derivatives, such as artemether, are highly
sensitive compounds, which require careful optimized produc-
tion processes for their manufacture. Due to robustness issues,
the manufacturing procedure of the reduction of artemisinin
with potassium borohydride to dihydroartemisinin was re-
investigated. The most important factor for obtaining optimal
yields is to ensure low levels of contamination of potassium
hydroxide in potassium borohydride. Application of a lower
reaction temperature, fast addition rate of potassium borohy-
dride, and careful control of the pH during the quench with
acid are further important parameters in guaranteeing a robust
process. In the redesign of the conversion of dihydroartemisinin
to artemether, the yield was increased, and dichloromethane
was replaced by the ecologically friendlier methyl acetate. A
robust manufacturing process for artemether is now at hand,
allowing the production of this important medicine reliably and
in good quality and yield.
Figure 1. The active ingredients for Coartem: artemether 1
and lumefantrine 2.
nonprofit basis to health authorities in malaria-endemic
developing nations since 2001. Production of Coartem, the
leading artemisinin-based combination therapy, has increased
from 100,000 treatments in 2002 to up to 65 million in 2006.
In view of the fast-growing demand, the improvement of
the manufacturing processes for both active ingredients, the
semi-natural product 1 and the fully synthetic tricyclic
compound 2 (Figure 1), were urgently needed. In this paper,
we present our results from the optimization of the manu-
facturing process for artemether (1). The following paper
will deal with the redesign of the manufacturing process for
lumefantrine (2).
Starting material for the manufacture of 1 is the natural
product artemisinin, 3, which is isolated from Artemisia
annua by extraction and purified by chromatography on silica
gel and crystallization.1a,b Artemisinin itself is also a powerful
antimalaria agent;2 however, due to the lactone function, it
shows relatively poor stability and has poor solubility
properties.3 It was therefore a breakthrough when it was
found that artemisinin, 3, could be reduced to dihydroarte-
misinin, 4, with sodium borohydride4 without affecting the
peroxide linkage. Treatment of 3 with methanol in the
presence of hydrochloric acid leads to 1 (Scheme 1).
The sequence as depicted in Scheme 1 was the basis for
the production process with which Novartis started its
production of artemether in autumn 2004.
1. Introduction
Coartem is a highly effective and well tolerated anti-
malarial that achieves cure rates of up to 95%, even in areas
of multidrug resistance. It is indicated for the treatment of
Falciparum malaria, the most dangerous form of malaria.
Coartem is the only prequalified, fixed-dose combination that
combines artemether, 1, an artemisinin derivative, and
lumefantrine, 2. This fixed-dose combination is of great
benefit to patients as it facilitates treatment compliance and
supports optimal clinical effectiveness.
Artemisinin, 3, is a compound derived from the sweet
wormwood plant and has been used for centuries in
traditional Chinese medicine to treat fever. An artemisinin-
based combination therapy is a combination of two or more
drugs (one of which is an artemisinin derivative) that have
different modes of action and different targets. Studies have
shown that using two or more drugs in combination has the
potential to delay the development of resistance in areas of
low transmission. Artemisinin-based combination therapies
in particular have been found to be highly effective treat-
ments for malaria. Their potential to delay resistance in areas
of intense transmission is under investigation.
2. Manufacture of Dihydroartemisinin, 4
The reduction of artemisinin 3 was performed by adding
potassium borohydride5 in portions to a suspension of 3 and
calcium chloride6 in methanol at 5 °C. After complete
Coartem was co-developed by Novartis in collaboration
with Chinese partners, and it is currently registered in 79
countries worldwide. Coartem has been extensively studied
in multicenter clinical trials involving more than 3000
patients. Under an agreement with the World Health Orga-
nization (WHO), Novartis has supplied Coartem on a
(1) (a) Tandon, S.; Kahol, A. P.; Jain, D. C.; Bhkuni, R. S.; Kumar, S. J. Sci.
Ind. Res. 2003, 62, 457. (b) Klayman, D. L.; Lin, A. J.; Acton, N.; Scovill,
J. P.; Hoch, J. M.; Milhous, W. K.; Theoharides, A. D. J. Nat. Prod. 1984,
47, 715.
(2) Clinical studies on the treatment of malaria with Qingaosu and its derivatives,
J. Tradit. Chin. Med. 1982, 2 (1), 45.
(3) Haynes, R. K.; Vonwiller S. C. Trans. R. Soc. Trop. Med. Hyg. 1984, 88,
Supplement 1, 23.
* To
whom
correspondence
should
be
addressed.
E-mail:
(4) Li, Y.; Yu, P.-L.; Chen, Y.-X.; Li, L.-Q.; Gai, Y.-Z.; Wang, D.-S.; Zheng,
Y. Kexue Tonbao 1979, 14, 667.
peter.fuenfschilling@balcab.ch.
336
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development
10.1021/op0602425 CCC: $37.00 © 2007 American Chemical Society
Published on Web 04/28/2007

reaction mixture.10 The methanol-induced decomposition
products of potassium borohydride had no effect either.11
Scheme 1. Manufacture of artemether 1
A
minor effect (-5% yield) from humidity was seen when
calcium chloride was used which had been exposed to air
for 24 h prior to use.
Thus far, with all these experiments we had gained
knowledge of how to prevent some yield losses, but we had
not yet found a deviation which would lead to a complete
failure. The situation changed when we ran the reduction in
the presence of potassium hydroxide.12 The effect of a
catalytic amount of potassium hydroxide on the degradation
rate of dihydroartemisinin, 4, proved to be dramatic. When
the reduction was run in the presence of only 1 mol % of
potassium hydroxide (relative to 3), a large part of 4 was
degraded.13 The isolated yield was below 10%, and the HPLC
fingerprints of the by-products in the mother liquor of this
laboratory experiment were almost identical with those of
failed batch 35. These observations suggest that small
amounts of potassium hydroxide have been the main root
cause for the complete failure.
2.2. Optimized Process Conditions for the Manufac-
ture of Dihydroartemisinin, 4. Having recognized the
extreme vulnerability of 4 towards small amounts of strong
bases, the further optimization strategy to make the reduction
reaction more robust became straightforward. Apparently,
the desired conversion 3 f 4 is in competition with the base-
induced degradation of 4. Even with a stringent limitation
of potassium hydroxide in potassium borohydride, an element
of risk would persist. Some traces of strong bases would
certainly not lead to a complete batch failure but would
nevertheless have a negative impact on the yield. As a
consequence, the reaction time has to be kept short, and the
reaction temperature kept as low as possible. In addition,
care has to be taken with the conditions of the quench with
hydrochloric acid. Consequently, by reducing the addition
time of potassium borohydride from around 10 h to below 6
h, adjusting the reaction temperature from 5 to 7 °C to 1-3
°C, and by keeping the pH during the acid quench at 4-6
(instead of 0-1), the whole process became very robust.
After reintroduction into production, an increase of the
average yield from originally 79% to 89% was observed. In
addition, due to the lower reaction temperature, the excess14
of potassium borohydride could also be reduced. Since a
consumption of 3, the excess of potassium borohydride was
destroyed by adding concentrated hydrochloric acid at 0-6
°C until a pH of 0.5-1 was reached. To precipitate the
product, water was added, and the pH of the reaction mixture
was adjusted to 5-6 by quenching the excess of hydrochloric
acid with aqueous potassium hydroxide. Filtration, washing
with water, and drying completed the manufacturing process
and led to pure dihydroartemisinin in 79% yield. Apart from
a somewhat above normal variation in the yield (74-84%),
no particular problems were encountered during the early
production batches in autumn 2004. However, from batch
20 on, the yield variation became larger, and after a dramatic
collapse7 in batch 35, the campaign was stopped, and a
program was initiated to rework the process.
2.1. Root Cause for Batch Failure. A batch failure to
such an extent as was experienced in batch 35 is a dramatic
event in a production plant, particularly with regard to the
loss of such a highly valuable starting material. At first, we
suspected a severe manipulation error; however, after an in-
depth analysis of the batch history, this possibility was ruled
out. The speculation of having used contaminated artemisinin
as starting material in batch 35 could also be excluded.
Consequently, we set up a program to systematically
investigate the influence of the various process parameters
on the yield by performing a series of scale-down experi-
ments for this reduction process. In parallel, the isolation of
the formed by-products in batch 35 was initiated as well
(section 2.3).
In view of the acid-sensitive hemiacetal moiety8a-d in 4,
studying the influence of the acidic quench conditions on
the yield seemed a good starting point for this investigation.
Indeed, when the quench was performed at 20 °C (instead
of 0-6 °C), the yield loss was 40%. Additionally, if the
amount of concentrated hydrochloric acid was increased from
1.5 to 2.3 equiv, a yield loss of 25% was observed.
Apparently, the quench conditions are important to obtain
an optimal yield, but they could not explain a complete
failure.9 Typical scale-up phenomenon such as the mode of
stirring showed only little or no effect at all. Also, no
negative effect was noted when iron salts were added to the
(10) It is assumed that the iron-catalyzed cleavage of the peroxy function of
artemisinin triggers its antiparasitic activity (see for instance Jefford, C.
W. Curr. Med. Chem. 2001, 8, 1803; Wu, Y. Acc. Chem. Res. 2002, 35,
255). This mechanism of action is in agreement with the cleavage of the
peroxy function when artemisinin is treated with FeCl2 in acetonitrile at 25
°C (Jefford, C. W.; Vincente, M. G. H.; Jacquier, Y.; Favarger, F.; Mareda,
J.; Millasson-Schmidt, P.; Brunner, G.; Burger, U. HelV. Chim. Acta 1996,
79, 1475). However, no decomposition was observed after 48 hours upon
treating artemisinin with 0.2 mol % FeCl2 in methanol at 5 °C.
(11) To investigate this hypothesis, potassium borohydride was completely
decomposed in methanol (leading to a pH around 9). After adding of
artemisinin 3 and calcium chloride, the reduction reaction was run under
standard conditions. No influence on the yield was noted.
(12) Potassium hydroxide is a “logical” impurity in potassium borohydride. The
latter compound is usually obtained by treating sodium borohydride with
ethanolic KOH and subsequent washing with ethanol.
(13) Besides 4, artemisinin, 3, is also sensitive towards strong base. However,
whereas catalytic amounts of a strong base are sufficient to cause significant
degradation in the lactol 4, molar quantities of base are needed in the case
of lacton 3.
(5) Compared to sodium borohydride, potassium borohydride is less hygroscopic
and therefore easier to handle on large scale.
(6) The reduction is also described with sodium borohydride without addition
of calcium chloride: Sing, C.; Tiwari, P. Ind. J. Chem. 2002, 41B, 2185.
(7) In this batch, the isolated yield dropped from an average yield of 79% to
only 2%.
(8) (a) Idowu, O. R.; Maggs, J. L.; Ward, S. A.; Edwards, G. Tetrahedron 1990,
46, 1871. (b) Yagen, B.; Pu, Y. M.; Yeh, H. J. C.; Ziffer, H. J. Chem. Soc.
Perkin Trans. I 1994, 843. (c) Baker, J. K.; Chi, H. T. Heterocycles, 1994,
38, 1497. (d) Imakura, Y.; Hachiya, K.; Ikemoto, T.; Yamashita, S.; Kihara,
M.; Kobayashi, S.; Shingu, T.; Milhous, W. K.; Lee, K. H. Heterocycles
1990, 31, 1011.
(9) Large deviations in temperature and amount of acid as applied in these
laboratory experiments could be ruled out for batch 35.
(14) The application of
a
lower reaction temperature leads to
a
slower
decomposition of potassium borohydride by methanol.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
337

Figure 3. Potential by-products in the conversion 4 f 1.
Scheme 2. Proposed pathway for the formation of the
diastereomers 13a,b
Figure 2. by-products isolated from the mother liquor of batch
35.
short reaction time is important and is dependent on the
dissolution rate of potassium borohydride in the reaction
mixture, the physical properties of this reagent are of
importance as well.15
3. Manufacture of Artemether 1
3.1. Conversion of Dihydroartemisinin to Artemether
in Dichloromethane. For the conversion of 4 to 1, the
applied procedure was improved. The reaction in dichlo-
romethane is performed in the presence of an excess of
methanol and concentrated hydrochloric acid at 28-32 °C.
After cooling to 12-15 °C, neutralizing with aqueous sodium
bicarbonate, and phase separation, most of the solvent is
distilled off, and crude 1 is precipitated18 by adding large
amounts of water. Filtration, washing with methanol/water,
and drying at 40 °C leads to crude 1 in 77% yield. The final
purification is accomplished by dissolving crude 1 in
methanol at 40-44 °C, precipitating with water at 40 °C,
and cooling to 5-10 °C. With this procedure, the overall
yield from 3 to 1 is 74%.
3.2. Improved Process Conditions for the Conversion
of Dihydroartemisinin to Artemether. In parallel to the
rework of the reduction step, we also wanted to improve
the ketalization 4 f 1. In particular we wanted to replace
dichloromethane by an ecologically more acceptable solvent
and to improve the yield by making the process more
selective, in particular by suppressing the formation of the
main by-product, 10R-epimer 5. Two other by-products, 9R-
epimer 619 and anhydroartemisinin 720 (Figure 3), were
formed only in small amounts; however, their depletion by
crystallization (in particular in the case of compound 6) was
more difficult.21
In retrospect, both the fluctuating yield in the first series
of production batches and the complete failure in batch 35
can now be understood. The most disastrous factor proved
to be the contamination of potassium borohydride with small
amounts of potassium hydroxide. If such a contamination
occurs in combination with a long reaction time, with a
reaction temperature at the higher end of the given range
and with coarse potassium borohydride particles, all condi-
tions are perfectly set up to lead to the observed failure of
batch 35.
2.3. Isolation of By-products from Batch 35. By using
preparative liquid chromatography followed by preparative
peak shaving recycling chromatography, the mayor degrada-
tion products (>80%) could be isolated from the mother
liquor of batch 35 as six pairs of diastereoisomers and their
1
structures elucidated with H NMR- and MS-spectroscopy
(Figure 2). The absolute configuration at carbon atom 3
(numbering see Scheme 1) could not be derived from the
NMR spectra.
The structures of the isolated by-products 10a-d and
11a,b are in agreement with earlier reports from the literature
on the base-catalyzed fragmentation of 4 under reductive
conditions.16 The acetals 12a-d were apparently derived
from 10a-d during the acidic workup. The formation of
13a,b (only in small amounts present in the mother liquor
of batch 35) may have started with the base-induced
formation of the monocyclic intermediate 14,17a-b followed
by the reduction of the various functional groups and
subsequent ring-closure reaction (Scheme 2).
As shown in Table 1, the best result was achieved when
the reaction was run in methyl acetate. In this solvent, the
ratio of 1/5 could be increased from 85:15 up to 89:11.
However, somewhat more of the undesired 9R-isomer 6 was
(18) In our plant, some of the first production batches failed to meet specifications
due to insufficient depletion of 10R-epimer 5 and 9R-epimer 6. To reach
the required quality, a recrystallization procedure had to be developed.
Furthermore, the precipitation and washing procedures had to be optimized
(for details, see Experimental Section).
(15) For details on the analytical specifications of potassium borohydride, see
Experimental Section.
(19) Pu, Y. M.; Ziffer, H. Heterocycles 1994, 39, 649.
(16) Sy, L. K.; Hui, S.-M.; Cheung, K. K.; Brown, G. D. Tetrahedron 1997, 53,
7493.
(17) (a) Zeng, M. Y.; Li, L. N.; Chen, S. F.; Li, G. Y.; Liang, X. T.; Chen, M.;
Clardy, J. Tetrahedron 1983, 39, 2941. (b) Shang, X.; He, C.; Zheng, Q.;
Yang, J.; Liang, X. Heterocycles 1989, 28, 421.
(20) Haynes, R. K.; Chan, H.-W.; Cheung, M.-K.; Chung, S. T.; Lam, W.-L.;
Tsang, H.-W.; Voerste, A.; Williams, I. D. Eur. J. Org. Chem. 2003, 11,
2098.
(21) Precipitation of artemether crude from methanol/water 55:45 mol/mol %
leads to >95% depletion of 5, 90% of 7, but only 40% of 6.
338
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Table 1. Reaction conditions and product distribution for
the conversion 4 f 1a
tion: min 98%. Calcium chloride water-free (purity g98%)
was purchased from Koma as granules.
5.2. Manufacture of Dihydroartemisinin (4), New
Process Conditions. In a 2500-L glass-lined reactor a
suspension containing 3 (117 kg, 415 mol), water-free,
granulated calcium chloride (15.6 kg, 141 mol), and methanol
(1290 L) was chilled to 0-4 °C with stirring. Potassium
borohydride (27 kg, 502 mol) was added in 3-kg portions.
Due to the evolution of hydrogen, the first portions were
added with great care.23 In our equipment, the following
parameters for the next addition of potassium borohydride
gave good results: internal temperature 0-4 °C, time
between portions: 10 min, gas flow: <9 m3/h, internal
pressure: max 1250 mbar. The progress of the reaction was
monitored by TLC. The reaction mixture was quenched when
the content of artemisinin was <2%. With hydroxide-free
borohydride the addition and reaction were complete within
3-6 h. The quench was performed by adding hydrochloric
acid 32% w/w (40 L) at -2 to +8 °C within 30 min, leading
to a pH of 4-6. Deionized water (1000 L) was added at
0-15 °C, raising the pH to 7-8. For efficient use of the
equipment train, the mixture was transferred to a 4000-L
glass-lined reactor, internal temperature was set to +2 to 15
°C, and deionized water (1200 L) was added. The product
was collected on a centrifuge and washed with deionized
water (10 × 240 L), dried in a shelf dryer for 9 h at 50 °C
to give 105 kg (89%) of dihydroartemisinin 4 as a white,
crystalline powder with a purity of 98% (HPLC).
temp HCl
1
5
6
7
solvent
methyl acetate
acetone
(°C) (equiv) (%) (%) (%) (%) 1/5
23
30
30
0.09 84.3 10.6
0.18 82.6 11.5
0.18 79.0 16.6
0.18 76.9 13.6
0.18 76.5 19.7
0.8 0.2 89:11
2.5 0.2 88:12
3.7 0.5 83:17
0.5 0.3 85:15
3.3 0.3 80:20
dichloromethane 30
dimethoxyethane 30
methanol
toluene
30
30
30
30
0.18 76.4 20.4 < 0.1 0.3 79:21
0.18 75.7 19.4
0.18 54.1 37.2
0.18 43.0 6.1
4.4 0.3 80:20
4.8 1.2 59:41
0.5 0.6 88:12
acetonitrile
tetrahydrofuranb
a Product distribution in the reaction mixture determined by HPLC; relative
response factors for 1/5/6/7 1.0:1.0:1.0:16.7. b Approximately 50% conversion
after 6 h.
also formed. Reduction of the temperature and of the amount
of acid reduces the formation of 6. Its depletion was assured
by precipitation of artemether in a mixture of methyl acetate,
methanol, and water, and the final purification by crystal-
lization from methanol and water. With these new process
conditions, the overall yield from 3 to 1 could be improved
from 74% to 77%.
4. Summary
The manufacturing procedure of the reduction of arte-
misinin with potassium borohydride to dihydroartemisinin
was thoroughly reworked. The most important factor for
obtaining optimal yields was to ensure low levels of
contamination of potassium hydroxide in potassium boro-
hydride. Application of a low reaction temperature, fast
addition rate of potassium borohydride, and careful control
of the pH during the quench with hydrochloric acid were
further parameters for guaranteeing a robust and optimal
process. In the conversion of dihydroartemisinin to arte-
mether, yield and quality were improved by an optimized
manufacturing procedure, including optional recrystallization.
Furthermore, dichloromethane could be replaced by methyl
acetate. With this solvent switch, a procedure is now at hand,
which is not only ecologically more friendly but also results
in an increased â:R ratio, giving rise to a somewhat better
overall yield.22
5.3. Isolation and Identification of By-products of
Batch 35. 5.3.1. Isolation of By-products 12a-d. Five
milliliters of mother liquor of batch 35 were filtered and
injected onto a Waters LC/MS system equipped with a
Micromass ZQ 2000 mass detector and a Waters Symmetry
semipreparative column (5 µm, 19 mm × 50 mm). Using
acetonitrile/water ) 4/6 (V/V) as the mobile phase at room
temperature and a flow rate of 20 mL/min, 10.5 mg of by-
products 12a,b (retention times: 11.5-11.75 min) and 23.5
mg by-products 12c,d (retention times: 12.5-13.26 min)
could be separated within one single run. The mass detector
in its atmospheric pressure ionization mode (AP+) turned
out to be superior in comparison to electrospray ionization
(ES+). The structures of all four isomers were confirmed
1
with H NMR- and MS-spectroscopy. Seventy-seven mil-
ligrams of the first eluting peak at 8.69 min turned out to be
a complex mixture of eight isomers which were separated
separately by preparative peak shaving recycling chroma-
tography.
5. Experimental Section
5.1. Starting Materials and Reagents. Artemisinin was
purchased from several sources in China and Vietnam; 100
mg gave a clear solution in 10 mL of acetonitrile/water (7/
3, v/v). Potassium borohydride was purchased from Finnish
Chemicals or Rohm & Haas and was used only within the
following specifications: appearance: fine, free-flowing
powder (no agglomerated, caked material); particle size:
median value approximately 50 µm, particles above 100 µm
less than 5% (laser diffraction); titration in sodium carbonate
solution with 0.1 M HCl: max 0.05 (basic contents) or max
0.35 (acidic contents) mL/g KBH4. Iodometric redox titra-
5.3.2. Isolation of By products 10a-d, 11a,b and 13a,b.
Evaporation residue (77 mg) of the first eluting peak (from
the previous separation) was dissolved in 5 mL of aceto-
nitrile, filtered, and injected onto a Kromasil C8 column (10
µm, 2.0 cm × 20 cm) while using the same LC/MS system
in its AP+ mode as mentioned above. Again acetonitrile/
water ) 4/6 (v/v) was used as mobile phase at room
temperature with a flow rate of 20 mL/min. By-product 10a,
eluting first, was removed directly, and the area of incomplete
separation was recycled. During the second run, by-product
10b was removed at the front of the elution profile, and
(22) The feasibility of the conversion 4 f 1 in methyl acetate has been
demonstrated on a kilogram scale. Since artemether will in future be
produced externally, this process was not implemented at Chemical
Production Novartis.
(23) With hydroxide-free borohydride, the reaction and gas evolution were
vigorous from the beginning.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
339

during the third run by-products 10c,d from the front and
by-product 13a from the end of the elution profile. The
remaining by-products, 11a, 11b, and 13b, were baseline
separated during the final recycling. The structures of all
of the critical by-products 5, 6, and 7, an optional recrys-
tallization procedure was applied: The wet filter cake was
dissolved in methanol (1100 L) and precipitated by adding
water (550 L) at 27-30 °C via a spray-ring device followed
by the same cooling, filtration, and washing procedure as
described above. Drying was performed under reduced
pressure in a 1600-L stainless steel double cone dryer for 4
h at 40 °C by applying a stream of nitrogen of 3000 L/h.
5.5. Preparation of Artemether Crude in Methyl
Acetate. Into a 1-L reactor equipped with a propeller stirrer,
a pH probe, and a distillation apparatus was placed a
suspension of dihydroartemisinin (50.0 g, 0.176 mol) in
methyl acetate (625 mL) which was then treated at 20 °C
with methanol (56.0 g, 1.75 mol) and aqueous hydrochloric
acid (2.1 g, 0.021 mol). The suspension was stirred for 4.5
h at 23 °C. The conversion of dihydroartemisinin was
monitored by TLC. The solution was cooled to 14 °C, and
a solution of sodium hydrogen carbonate (6 g) in water (75
mL) was added. The mixture was stirred for 10 min, and
phase separation was performed at pH 6-8. The organic
phase (∼750 mL) was concentrated (250-200 mbar) to a
volume of ∼160 mL. The residue was diluted with methanol
(185 mL) and concentrated in vacuo (200-100 mbar) to a
volume of ∼160 mL. The suspension was heated to 29 °C
and diluted with methyl acetate (31 mL) and methanol (94
mL). The resulting solution was filtered. The reactor and
the filter were washed with methanol (50 mL). The solution
was treated at 29 °C with water (140 mL) over 2 h. The
resulting suspension was cooled in 2 h to 2 °C and stirred at
this temperature for at least 1 h. The product was filtered
and washed with cold methanol/water 1:1 (100 mL) and with
cold water (2 × 50 mL). Drying (∼10 mbar/ 40 °C) afforded
artemether crude (42.0 g, 80%) as sand-like, colorless
crystals.
1
by-products were confirmed with H NMR- and MS spec-
troscopy. As mentioned in section 2.3, it is at present not
possible to distinguish the stereochemistry at carbon 3 (even
for the isolated isomers); the NMR data are therefore not
disclosed. Complete structure elucidation will only be
possible after applying further structure elucidation technolo-
gies.
5.4. Manufacture of Artemether Crude in Dichlo-
romethane. 4 (160 kg, 562 mol) was loaded into a 2500-L
glass-lined reactor. Dichloromethane (2000 L) was added,
and the suspension was stirred at a moderate speed. Methanol
(150 L) and concentrated hydrochloric acid (9 L) were added.
The now almost clear solution was heated to 28-32 °C
(applying a target jacket temperature of 45 °C), and the
solution was stirred at 28-32 °C for 3 h. Reaction progress
was monitored by TLC (requirement: g2% of 4). The
reaction mixture was cooled to 11-15 °C, and a solution of
sodium bicarbonate (10 kg) in water (240 L) was added. The
organic phase was transferred to a 4000-L stainless steel
reactor, and the aqueous phase was extracted with dichlo-
romethane (2 × 160 L). The combined organic phases were
concentrated to ∼400 L under reduced pressure at 18-23
°C. Methanol (550 L) was added. The mixture was trans-
ferred to a 1600-L stainless steel reactor, and the reactor and
pipes were flushed with dichloromethane (50 L). The mixture
was once more concentrated to ∼400 L under reduced
pressure at 20-30 °C. This procedure was repeated by
addition of 600 L of methanol and subsequent distillation to
∼400 L volume. After addition of methanol (500 L) at 20
°C, the solution was clear-filtered, and the reactor and pipes
were flushed with methanol (160 L) into the precipitation
reactor (4000-L stainless steel). The solution was heated to
29 °C using a target jacket temperature of 45 °C. Artemether
crude was precipitated by the addition of water (620 L).24
To reach the required quality of pure artemether, water was
added at a constant rate within ca. 2 h via a lance especially
designed for this purpose. Nitrogen pressure was applied to
spray the water onto the surface of the methanol solution,
which was kept at 27-30 °C throughout the addition of
water. Initially, the set point in our procedure was 520 L of
water. Using less than 620 L of water for precipitation gave
the same quality; using over 640 L resulted in the precipita-
tion of by-products. During process optimization, 620 L
proved to be the optimal amount of water to obtain the
highest yields.25 After addition of water, the suspension was
stirred for 15 min, cooled to 2 °C, and stirred for 1 h at 2
°C. After filtration on a centrifuge, the filter cake was washed
with a 1:1 (v/v) mixture of methanol and water (320 L), and
with water (2 × 160 L). Typical impurity levels of water-
wet product: 5 < 0.1% and 7 ) 0.2%. The average yield
using this optimized manufacturing procedure was 77.1%.
If the wet product did not comply with respect to the content
5.6. Crystallization of Artemether Crude. Artemether
crude (150 kg) was dissolved in methanol (650 kg) at 40-
44 °C. After clear filtration and rinsing with methanol (190
L), water (175 kg) was added at 40 °C in 30 min, and the
solution was seeded in three portions with a total of 3 kg of
artemether pure. Again at 40 °C, water was added (710 kg)
in ca. 3 h. The suspension was cooled within 2-3 h to 5-10
°C. Filtration was performed on a centrifuge, and the product
was washed in three portions with a precooled 1:1 methanol/
water mixture. After drying, 144 kg (96%) of pure artemether
1 was obtained.
Acknowledgment
We thank M. Baysang, B. Betschart, Ch. Bury, M. Gasser,
R. Humair, F. Langhart, D. Roth, M. Scha¨fer, M. Scheideg-
ger, and A. Smogavec for their enthusiastic support of this
work, and L. Oberer and K. Akyel for NMR measurements
and interpretation. We also thank Jerh Collins for his
encouraging support during the critical phases in the produc-
tion plant and, in particular, Thomas Kuhn, Global Project
Leader for Coartem, for his highly appreciated assistance
during the whole duration of the project.
(24) Since four out of the first six batches failed to meet the specifications due
to high levels of by-products (up to 2.3% of 5 and up to 0.67% of 7), amount
and addition mode of water had to be optimized.
Received for review November 16, 2006.
OP0602425
(25) In this optimization work we found also that seeding is not necessary.
340
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development